超声乳腺肿瘤图像的自动分割  

Automatic boundary extraction of ultrasonic breast tumor image

作　　者：刘红[1] 张欣[1] 吕丹[1] 程海凭[1] 

机构地区：[1]上海医疗器械高等专科学校医用电子信息系,上海200093

出　　处：《计算机工程与设计》2010年第15期3537-3538,F0003,共3页Computer Engineering and Design

基　　金：上海高校选拔培养优秀青年教师科研专项基金项目(SLG05052)

摘　　要：为解决传统Snake模型对初始轮廓敏感和凹陷边界提取困难的问题,分别提出了双阈值分割算法和Snake模型的改进方案。通过双阈值分割算法与形态学运算、滤波技术的综合应用,获得靠近边缘的初始轮廓。采用改进的Snake模型使初始轮廓跟踪到实际边界。由于对模型的能量定义做了调整,凹陷边界也可被准确跟踪。通过临床采集的20例乳腺图像肿瘤边缘的提取和分析,结果表明,该方法能有效提取出肿瘤边缘,实现超声肿瘤的自动分割。To solve the problem of high requirements on initial contour for traditional Snake model,and make the extraction of cupped boundary more accurate,a segment algorithm of dual-threshold and an improved Snake model are proposed respectively.Firstly,a nearby contour is obtained,based on the algorithms of dual-threshold segmentation,morphological operation and filtering technique.Then,a more accurate contour is tracked by the improved Snake model.Since the energy definition of this model is adjusted,cupped boundary can also be traced accurately.Twenty experiments on boundary extraction from breast ultrasonic images have proved that this proposed method can effectively extract the boundary of breast tumor,and realize the auto-segmentation of ultrasonic tumor image.

关 键 词：乳腺肿瘤 边缘提取 低高中滤波器 阈值分割 动态轮廓模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]